Research On The Load Forecasting Method Based On Time Series Arrangement And Artificial Intelligence

The accuracy of power load forecasting directly affects the operation and planning of power grid institutions.Accurate power load forecasting has important reference value for power system operation.Due to the existence of problems such as the non-stationary characteristics of power load data,the instability of forecasting models and the difficulty in mining potential information of load data,the current short-term power load forecast accuracy is still difficult to meet the needs of real-time power grid dispatch.In order to improve the accuracy of short-term power load forecasting,the short-term power load is forecasted from the power load data processing,characteristic variable analysis and construction of forecasting models.The change of the load sequence not only has a non-linear periodic regular change;at the same time,it is affected by external factors,showing a certain degree of randomness and non-linear characteristics.In order to explore the potential change characteristics of the original load data,a variational modal decomposition(VMD)technology is introduced to decompose the original load data.According to the number of different modes and the value of the penalty factor,the VMD decomposition effect will be greatly affected.The Sparrow Search Algorithm(SSA)is used to optimize its and parameters,and a load decomposition prediction model based on SSA-VMD is constructed.Variational modal decomposition(VMD)is used to decompose the original load data into modal components with different characteristics,and a load forecasting model is established according to the characteristics of each modal component.The mutual information method is used to analyze the correlation between the characteristic variables and the load,and the characteristic variables that have a greater impact on the load are selected as the input characteristic indicators of the prediction model.And according to the characteristics of each modal component,the characteristic index that can reflect the change characteristics of each component is selected as the input variable of the prediction model.Aiming at the time series and nonlinear characteristics of load,according to the characteristics of each modal component,this paper establishes the load forecasting model of least square support vector machine(LSSVM)and long short-term memory neural network(LSTM).Due to the complex changes in the load sequence and the many factors affecting the load,the single-layer LSTM has insufficient data learning performance and adaptability.A two-layer two-way LSTM neural network prediction model is constructed,and the Sparrow algorithm(SSA)is used to analyze the two-layer two-way LSTM The parameters in the LSSVM model are optimized to improve the prediction performance of the model.The model of DBi LSTM without VMD decomposition is selected to compare the prediction results.The experimental simulation results show that the prediction after the VMD decomposition and the real load change curve have a higher degree of fit.In order to further verify the superiority of the proposed prediction model,VMD-SSA-DBi LSTM,VMD-SSA-LSSVM and VMD-SSA-LSTM compare the prediction results to prove that the VMD-SSA-DBi LSTM-LSSVM power load prediction model has good prediction accuracy And stability.